Much of the scientific process involves searching. But rather than continue to rely on the luck of discovery, Google DeepMind has engineered a more efficient AI agent that mines complex spaces to facilitate scientific breakthroughs. Sarah Guo speaks with Pushmeet Kohli, VP of Science and Strategic Initiatives, and research scientist Matej Balog at Google DeepMind about AlphaEvolve, an autonomous coding agent they developed that finds new algorithms through evolutionary search. Pushmeet and Matej talk about how AlphaEvolve tackles the problem of matrix multiplication efficiency, scaling and iteration in problem solving, and whether or not this means we are at self-improving AI. Together, they also explore the implications AlphaEvolve has to other sciences beyond mathematics and computer science.

Sign up for new podcasts every week. Email feedback to [email protected]

Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @pushmeet | @matejbalog

Chapters:

00:00 Pushmeet Kohli and Matej Balog Introduction

0:48 Origin of AlphaEvolve

02:31 AlphaEvolve’s Progression from AlphaGo and AlphaTensor

08:02 The Open Problem of Matrix Multiplication Efficiency

11:18 How AlphaEvolve Evolves Code

14:43 Scaling and Predicting Iterations

16:52 Implications for Coding Agents

19:42 Overcoming Limits of Automated Evaluators

25:21 Are We At Self-Improving AI?

28:10 Effects on Scientific Discovery and Mathematics

31:50 Role of Human Scientists with AlphaEvolve

38:30 Making AlphaEvolve Broadly Accessible

40:18 Applying AlphaEvolve Within Google

41:39 Conclusion

When will fully autonomous vehicles see widespread adoption? According to Applied Intuition, that future is closer than you may think. Applied Intuition’s CEO, Qasar Younis, and CTO, Peter Ludwig, talk with Elad Gil about how now is the best time to both work on self-driving vehicle technology and monetize it. Qasar and Peter discuss the advantages of developing their own OS in-house for their autonomous applications, self-driving technology’s potential to drive re-shoring of vehicle manufacturing to the United States, and how best to gauge the bar for safety in autonomous systems. Plus, they explore how self-driving technology may reshape the designs of not only vehicles, but cities themselves.

Sign up for new podcasts every week. Email feedback to [email protected]

Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @qasar | @AppliedInt

Chapters:

00:00 Qasar Younis and Peter Ludwig Introduction

01:28 A Primer on Applied Intuition

11:08 Applied Intuition’s Customers

12:04 Impact of Chinese Vehicles Manufacturers

15:44 EV Policies in the European Market

20:49 Can Robotics and Automation Re-Shore Vehicle Manufacturing?

21:53 Training Models for Autonomous Vehicles

26:41 Gauging the Bar for Autonomous Vehicles Safety

32:03 Timeline for Large-Scale Autonomous Vehicle Adoption

36:28 Rethinking Urban Design for Autonomous Vehicles

38:47 How Applied Intuition Uses AI for Tooling and OS

42:09 Designing for User Experience

43:31 Applied Intuition’s Hiring Strategy

45:01 Conclusion

What happens when you give AI researchers unlimited compute and tell them to compete for the highest usage rates? Ben Mann, Co-Founder, from Anthropic sits down with Sarah Guo and Elad Gil to explain how Claude 4 went from "reward hacking" to efficiently completing tasks and how they're racing to solve AI safety before deploying computer-controlling agents. Ben talks about economic Turing tests, the future of general versus specialized AI models, Reinforcement Learning From AI Feedback (RLAIF), and Anthropic’s Model Context Protocol (MCP). Plus, Ben shares his thoughts on if we will have Superintelligence by 2028. 

Sign up for new podcasts every week. Email feedback to [email protected]

Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @8enmann

Links: 

ai-2027.com/ 

Chapters:

00:00 Ben Mann Introduction

00:33 Releasing Claude 4

02:05 Claude 4 Highlights and Improvements

03:42 Advanced Use Cases and Capabilities

06:42 Specialization and Future of AI Models

09:35 Anthropic's Approach to Model Development

18:08 Human Feedback and AI Self-Improvement

19:15 Principles and Correctness in Model Training

20:58 Challenges in Measuring Correctness

21:42 Human Feedback and Preference Models

23:38 Empiricism and Real-World Applications

27:02 AI Safety and Ethical Considerations

28:13 AI Alignment and High-Risk Research

30:01 Responsible Scaling and Safety Policies

35:08 Future of AI and Emerging Behaviors

38:35 Model Context Protocol (MCP) and Industry Standards

41:00 Conclusion

In this episode of No Priors, Sarah and Elad are joined by Dr. Fei-Fei Li, AI pioneer, co-director of Stanford’s Human-Centered AI Institute, and founder of World Labs. Fei-Fei shares why she’s building at the intersection of embodiment and intelligence, and what today’s AI systems are still missing. From the early days of ImageNet to her vision for the next generation of robotics, she unpacks the human and technical motivations behind World Labs. They also discuss the challenges of 3D world modeling, her approach to building exceptional teams, and the special qualities that have led her students like Andrej Karpathy to make major breakthroughs.

Show Notes:

0:00 Why and what Dr. Fei-Fei Li is building

3:00 World models at World Labs

6:44 Missing gaps in the AI future

9:16 Robotics and physical intelligence

16:15 Greatest challenges of 3D

19:08 Fei-Fei’s work in PhD in ImageNet

23:05 Special moments in Dr. Li's career

29:33 Building teams

32:05 Human-centered AI

In this episode of No Priors, Sarah and Elad unpack the current state of the AI market - whether it’s consolidating, what’s enabling or blocking key mergers, and where the most promising untapped opportunities lie, particularly in biotech. They also explore the rise of world models and how AI’s novel methods for understanding complex systems may ultimately reshape how humans approach discovery and problem-solving. 

Show Notes:

0:00 Is the AI market consolidating into clear winners? + the physics of the current landscape

7:01 Why more companies don’t merge (even when it makes sense) 

10:09 Exploring biotech’s biggest commercial opportunities and the challenges founders face 

17:14 Building world models 

21:34 How AI is expanding the way humans reason, design, and evolve systems

Arvind Jain joins Sarah and Elad on this episode of No Priors. Arvind is the founder and CEO of Glean, an AI-powered enterprise search platform. He previously co-founded Rubrik and spent over a decade as an engineering leader at Google. In this episode, Arvind shares how LLMs are transforming enterprise search, why most tools in the space have failed, and the opportunity to build apps powered by internal knowledge. He discusses how much customization is still needed on top of foundation models, what made building Glean uniquely challenging compared to Arvind’s previous ventures, and what’s next for the company.

Sign up for new podcasts every week. Email feedback to [email protected]

Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @jainarvind

Show Notes:

0:00 Introduction

0:58 How LLMs are changing search

2:05 Building out Glean’s platform

5:09 Why most search companies failed

8:41 Out of the box vs. bespoke models 

10:26 Creating apps on top of internal knowledge

15:34 User behaviors & insights 

19:11 Unique challenges of building Glean 

21:51 Product-led growth vs. enterprise sales

25:00 Succeeding in traditionally bad markets 

27:08 What Glean is excited to build next

On this episode of No Priors, Sarah talks to Luis von Ahn, founder and CEO of Duolingo, the world’s most popular education app with over 116 million monthly users and a market cap of approximately $17 billion. Controversially, it has recently committed to being “AI-first.” They discuss why motivation is the biggest challenge in education, how Duolingo harnesses game mechanics and behavioral insights to keep learners engaged, and the company’s efforts to leverage AI to personalize education at scale. Luis also shares thoughts on the Duolingo brand, courses beyond language (chess and math), and the broader impact of AI on content creation.

Sign up for new podcasts every week. Email feedback to [email protected]

Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @LuisvonAhn

Links:

Show Notes:

0:00 Introduction

4:01  Optimizing learning behavior through tech

11:20 Adopting AI at Duolingo

17:25 AI’s threat to content companies

18:34 An unhinged corporate brand

21:28 How do people learn?

25:16 What people misunderstand about Duolingo?

26:24 How AI is transforming learning at scale

30:28 Leveraging AI across the business

This week on No Priors, Elad and Sarah sit down with Eric Mitchell and Brandon McKinzie, two of the minds behind OpenAI’s O3 model. They discuss what makes O3 unique, including its focus on reasoning, the role of reinforcement learning, and how tool use enables more powerful interactions. The conversation explores the unification of model capabilities, what the next generation of human-AI interfaces could look like, and how models will continue to advance in the years ahead.

Sign up for new podcasts every week. Email feedback to [email protected]

Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @mckbrando | @ericmitchellai

Show Notes:

0:00 What is o3?

3:21 Reinforcement learning in o3

4:44 Unification of models

8:56 Why tool use helps test time scaling

11:10 Deep research

16:00 Future ways to interact with models

22:03 General purpose vs specialized models

25:30 Simulating AI interacting with the world

29:36 How will models advance?

On this episode of No Priors, Sarah sits down with Isa Fulford, one of the masterminds behind deep research. They unpack how the initiative began, the role of human expert data, and what it takes to build agents with real-world capability and even taste. Isa shares the differences between deep research and OpenAI’s o3 model, the challenges around latency, and how she sees agent capabilities evolving. Plus, OpenAI has announced that deep research is free for all US users starting today.

Sign up for new podcasts every week. Email feedback to [email protected]

Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @IsaFulf

Show Notes:

0:00 Deep research’s inception & evolution

6:12 Data creation 

7:20 Reinforcement fine-tuning

9:05 Why human expert data matters

11:23 Failure modes of agents

13:55 The roadmap ahead for Deep Research

18:32 How do agents develop taste? 

19:29 Experience and path to building a broadly capable agent

22:03 Deep research vs. o3

25:55 Latency

27:56 Predictions for agent capabilities

This week on No Priors, Sarah and Elad are joined by Josh Goldman, cofounder and president of KoBold Metals. KoBold is using AI to transform how we discover critical minerals like lithium and cobalt, making the exploration process faster, more precise, and more scalable than traditional methods. In this episode, Josh explains how KoBold is rethinking the fundamentals of mineral exploration by combining unique datasets, scientific modeling, and predictive algorithms. They dive into the company’s driving philosophy and technical approach, how they validate underground hypotheses, and why regulatory knowledge and a localized approach are crucial. Josh also discusses what success looks like in exploration today and the scarcity of world-class deposits.

Sign up for new podcasts every week. Email feedback to [email protected]

Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @KoBold_Metals

Show Notes:

0:00 Introduction

0:29 KoBold Metals

3:14 Using unique datasets

6:20 Traditional methods of lithium exploration

8:38 Regulatory vs. rarity constraints

13:40 Technical approach

16:25 Validating hypotheses

23:56 Redefining success in mineral exploration

25:44 Scarcity of good projects and deposits

32:44 Philosophy behind prediction

36:46 KoBold’s origin story